We previously studied a sample of patients with nonmelancholic depression and attempted to subtype their depression on the basis of etiological factors and current clinical features, analyzed individually and together. We now examine the validity of the three contrasting analytic approaches by examining outcome trajectories in a prospective study. At follow-up we reinterviewed 67% of our initially assessed nonmelancholic subjects, with progress over the 12-month interval assessed across a range of outcome parameters. The "anxious depressed" groups identified across all three differing analytic strategies consistently had the poorest 12-month outcome. However, varying definitions of this group influenced quite differing trajectories, with the worst outcome being for those defined on the basis of clinical features. Study results strongly suggest that patients presenting with a nonmelancholic depressive disorder and who acknowledge a significant number of concurrent anxiety symptoms are likely to have a distinctly poor medium-term outcome. Thus, clinical definition does have predictive utility in such a group.